Sequence detection and parameter estimation over the wireless multichannel based on fractional and higher order statistics

The demand on the wireless technologies is explosive with the innovation on the communication systems. The high speed transmission with the large number of users in the capacity-limited channel requires clever and efficient detection algorithms to keep the reliable connection. In this dissertation, we propose new algorithms for the sequence detection in an impulsive channel and for the blind multichannel identification and deconvolution in the presence of the Gaussian noise.; First, we present sequence estimation algorithms based on alpha stable distribution in the impulsive noise channel. We investigate the applicability of the Iterative Reweighted Least-Square (IRLS) algorithm and newly developed IRLS with decision feedback (DF) algorithm. We also develop a Cauchy-based maximum likelihood sequence estimation (MLSE) with reduced complexity for the multiuser detection called Cauchy survivor feedback MLSE (SF-MLSE) and assess the performance in wireless application problems through simulations.; Second, we develop a new method for the blind multichannel identification and the blind deconvolution algorithm based on higher-order statistics and eigenvector decompositions. We show the existence of a feasible solution for the blind multi-channel identification based on the fourth order cumulant only. The main emphasis is to put on a robust algorithm for the blind multichannel deconvolution applied to the source separation problems. For that purpose, we improvise the new separation principle and develop the Blind Multi-channel Generalize Eigenvector (BMGE) algorithm. The proposed separation principle is shown to separate source signals up to the ambiguity of the scalar multiplicity. The analytic investigation shows that the converging equalizer leads to the near MMSE performance and simulation result shows the excellent recovery of the original source signals.; Finally, we propose the unified approach to deconvolute the periodically time-varying multi-channel with multi-rate signaling. To do so, the slight modification of the BMGE algorithm is found to be good solution to this problem. As an application, it is applied to the multi-user detection of the multi-rate WCDMA signals and achieves the satisfactory performance in terms of the BER at high signal to noise ratio (SNR).